Blender Git Loki

svn merge -r41926:41932 ^/trunk/blender

merge mempool changes from bmesh (adds mempool iterator).

remove NULL check for the pool, trunk doesnt have this, caller better check.

Fix #29260: Missing "Extend" parameter for Border Select

Added "Extend" flag to border select operators for editors:
- UV Editor
- Sequencer
- NLA
- Info Space
- Graph Editor
- File Browser
- Clip Editor
- Action Editor
- Channels and markers regions

Can be used for custom keymaps.

better alignement for BLI_mempool struct

svn merge -r41899:41926 ^/trunk/blender. also sync mempool with trunk and move BLI_mempool_alloc out of mempools header where it was inlined

Dynamic Paint:
* Wave simulation speed doesn't anymore depend on surface size, but uses relative distances instead. This change will likely change simulation behavior on existing saves, but can be easily tweaked back using the "Wave Speed" parameter.
* Added a new wave brush type, "Depth Change". It uses the change of brush intersection between frames, giving a better looking "wake" for moving objects. It also doesn't leave any "dent" to the surface while remaining still.

replace macros with math functions

quiet compiler warnings for BLI_array defines, split BLI_array_append into BLI_array_append / BLI_array_append_r, the latter returning the new array location.

minor edits, move mempool stack vars into the nested scope when they aren't likely to be used, also formatting edits (was quite un-blender like).

minor cleanup
- remove / comment unused python vars
- replace mul_v3_fl(somevec, -1.0f); with negate_v3(somevec);

Cycles: cmake build system tweak, might solve compile issue.

Fix #29292: cycles not loading linked image datablocks correctly.

Fix #29293: NLA Strip modifiers don't survive save/reload. One letter
can make a big difference :)

edit the _BLI_array_growone macro so it doesnt give the error that values are calculated but not used.

Fix #29093: world zenith up and down texture influence were not working
correct. These were decoupled from horizon influence for 2.5, but not
actually used in the render engine.

tag index arrays as dirty when any element is removed, also check validity on operator init/exit (check was missed in some cases before)

Fix #29160: material node "Texture" didn't use default texture coordinates
anymore when nothing was connected to the socket.

Fix #29243: unlink texture datablock from material node in texture properties
could crash, RNA pointer from context was invalid.

Pep8 changes for motion tracking py scripts